Method and System for Automated Analysis of Nuclear Magnetic Resonance Spectra

1. A method for analyzing an MR spectrum or a nuclear magnetic resonance spectrum, the method using at least two expert systems having chemical and MR-expert analysis tools, the method comprising the steps of: a) initially characterizing all data in the MR spectrum as non-interpreted data; b) feeding, following step a), the non-interpreted data to an input of a first expert system; c) analyzing, in the first expert system, a spectral content of the non-interpreted data of step a) with regard to first specific spectral features thereof which the first expert system is structured to assess and reassigning that non-interpreted data into newly interpreted data and newly non-interpreted data, thereby associating a probability weighting to both the newly non-interpreted and the newly interpreted data; d) feeding the newly interpreted data and the newly non-interpreted data to a controller, the controller structured to evaluate progress in reassigning non-interpreted data into interpreted data and deciding, in dependence on that progress, either to stop an analyzing process or to feed the interpreted and the non-interpreted data back to the input of the first expert system or to an input of a second expert system; e) in the event that the controller decides to feed the interpreted and non-interpreted data back to the first expert system, re-analyzing, in the first expert system, a spectral content of the non-interpreted and interpreted data with regard to the first specific spectral features thereof and re-assigning the non-interpreted data and interpreted data into newly interpreted data and newly non-interpreted data, thereby associating a new probability weighting to both the interpreted data and the non-interpreted data; f) in the event that the controller decides to feed the interpreted and non-interpreted data to the second expert system, re-analyzing, in the second expert system, a spectral content of the non-interpreted and interpreted data with regard to second specific spectral features thereof which the second expert system is structured to assess and re-assigning the interpreted data and the non-interpreted data into newly interpreted data and newly non-interpreted data, thereby associating a new probability weighting to both the non-interpreted and the interpreted data; g) repeating step d); and h) in case of stopping the analyzing process, either sending out a statement that the method could not interpret the data or sending out a resulting information comprising a pair of interpreted and non-interpreted data containing a highest information content compared to other pairs of interpreted and non-interpreted data, wherein a new probability weighted pair of interpreted and non-interpreted data is fed at least twice to one of the first and the second expert systems.

2. The method of claim 1 , wherein a plurality of new probability weighted pairs of interpreted and non-interpreted data are generated at an output of a respective expert system.

3. The method of claim 1 , wherein probability weighted extracted spectral information are stored as new interpreted data only in case an associated probability factor exceeds a previously defined threshold.

4. The method of claim 1 , wherein for a first assignment of probabilities to extracted spectral information from the MR-spectrum, an initial guess for these probabilities are made using chemical and MR-expert knowledge resident in the first expert system.

5. The method of claim 1 , wherein said first expert systems is a peak identifier and generation of a new probability weighted pair of interpreted and non-interpreted data is carried out by a peak picking routine, wherein peak hypotheses are made and stored as new interpreted data.

6. The method of claim 5 , wherein extracted spectral information comprises chemical shift information, J-coupling information, integral information, and/or line shape information.

7. The method of claim 1 , wherein said second expert systems is a multiplet identifier, wherein generation of a new probability weighted pair of interpreted and non-interpreted data is carried out by grouping peak clusters into possible multiplets and assigning probabilities to these multiplet hypotheses, wherein the multiplet hypotheses are stored as new interpreted data.

8. The method of claim 7 , wherein a number of atoms is assigned to the multiplet hypotheses.

9. The method of claim 7 , wherein extracted spectral information is chemical shift information, shape information, integral information, multiplet patterns, couplings between different multiplets, structural information, information about a number of atoms that give rise to certain multiplets in the spectrum, information about spectral artifacts and/or information about spectral signals which do not originate from a main compound.

10. The method of claim 1 , wherein a new pair of interpreted and non-interpreted data is fed to an input of another expert system which is a structure analyzer and spectra matching system, wherein the structure analyzer and spectra matching system predicts spectral parameters from a given expected molecular structure, assigns probability factors to predicted spectral parameters, generates a new probability weighted pair of interpreted and non-interpreted data by analyzing consistency of the new interpreted and non-interpreted data with the predicted spectral parameters with their assigned probability factors, and feeds the new pair of interpreted data and non-interpreted data to the controller.

11. A system for analyzing an MR spectrum or a nuclear magnetic resonance spectrum, the system using at least two expert systems having chemical and MR-expert analysis tools, the system comprising: means for initially characterizing all data in the MR spectrum as non-interpreted data; means for feeding the non-interpreted data to an input of a first expert system; means for analyzing, in the first expert system, a spectral content of the non-interpreted data with regard to first specific spectral features thereof which the first expert system is structured to assess and reassigning that non-interpreted data into newly interpreted data and newly non-interpreted data, thereby associating a probability weighting to both the newly non-interpreted and the newly interpreted data; means for feeding the newly interpreted data and the newly non-interpreted data to a controller, the controller structured to evaluate progress in reassigning non-interpreted data into interpreted data and deciding, in dependence on that progress, either to stop an analyzing process or to feed the interpreted and the non-interpreted data back to the input of the first expert system or to an input of a second expert system; means, in the event that the controller decides to feed the interpreted and non-interpreted data back to the first expert system, for re-analyzing, in the first expert system, a spectral content of the non-interpreted and interpreted data with regard to the first specific spectral features thereof and re-assigning the non-interpreted data and interpreted data into newly interpreted data and newly non-interpreted data, thereby associating a new probability weighting to both the interpreted data and the non-interpreted data; means, in the event that the controller decides to feed the interpreted and non-interpreted data to the second expert system, for re-analyzing, in the second expert system, a spectral content of the non-interpreted and interpreted data with regard to second specific spectral features thereof which the second expert system is structured to assess and re-assigning the interpreted data and the non-interpreted data into newly interpreted data and newly non-interpreted data, thereby associating a new probability weighting to both the non-interpreted and the interpreted data; means for iteratively passing the non-interpreted data and the interpreted data from the first and second expert systems to the controller; and means, in case of stopping the analyzing process, for either sending out a statement that the data could not be interpreted or for sending out a resulting information comprising a pair of interpreted and non-interpreted data containing a highest information content compared to other pairs of interpreted and non-interpreted data, wherein a new probability weighted pair of interpreted and non-interpreted data is fed at least twice to one of the first and the second expert systems.

12. The system of claim 11 , wherein the system further comprises an initializer for assigning data of an MR spectrum to an interpreted data category or to a non-interpreted data category, wherein the initializer has an input for receiving MR data.

13. The system of claim 12 , wherein the first expert system is a peak identifier and the second expert system is multiplet identifier.

14. The system of claim 13 , further comprising a third expert system which is a structure analyzer and spectra matching system, wherein the structure analyzer and spectra matching system has a further input for receiving expected molecular structure information.